As is well known, NPN transistors are preferred for implementing integrated circuitry of the bipolar type on semiconductor substrates, on account of their DC and AC amplifications being definitely better than those of PNP transistors.
Another disadvantage of PNP bipolar transistors is their close limitations in high frequency applications.
Despite all this, the inclusion of both NPN and PNP transistors on a common semiconductor substrate is still a necessity where suitable bias circuits, current mirror circuits and/or load devices for gain stages are to be provided.
The formation of PNP transistors on a semiconductor substrate is effected concurrently with that of NPN transistors, without any additional implanting or masking steps.
It is for this reason that it has become common practice to form PNP bipolar transistors of the so-called lateral type such that they can be compatible with the process flows adopted to fabricate NPN bipolar devices.
In this respect, FIG. 1 shows an enlarged cross-sectional view, taken on a vertical plane, of a sidewall PNP device 6 formed on a P-doped semiconductor substrate 1.
The following are successively deposited onto this substrate 1: a first buried layer 2 doped N+ to form the base region of the transistor, and a second layer 3, doped N, which constitutes the active area of the PNP device to be.
Thereafter, a selective diffusion of P-type dopants is realized in the active area 3 to define a central emitter region 4 surrounded by two opposite collector regions 5.
This solution has been widely used heretofore, to the point that a text, "Design and Realization of Bipolar Transistors", Peter Ashburn, page 157, gives it as the principal configuration for a lateral PNP device that can ensure of a good current gain.
It should be noted that such lateral PNP transistors have an intrinsic current loss of about 3% compared to collector current; these currents can also be explained theoretically by having reference to the diagram in FIG. 2 of an equivalent electric circuit of a lateral PNP transistor.
It can be seen in this figure that the lateral PNP device, designated Q.sub.1 in the equivalent circuit, is connected to a pair of parasitic PNP transistors Q.sub.2 and Q.sub.3 having their respective emitter regions connected to the emitter and collector regions of the transistor Q.sub.1.
In addition, these parasitic devices Q.sub.2 and Q.sub.3 have their collector regions connected to the semiconductor substrate, and their base regions in common with the base of Q.sub.1.
This equivalent electric diagram shows that the parasitic currents are mainly attributable to the second parasitic device Q.sub.2 draining toward the substrate some of the current being injected by the emitter of Q.sub.1, thereby lowering the efficiency of lateral emission.
Similar considerations apply to the third parasitic device Q.sub.3, which contributes instead to lowering the collection efficiency of the collector of Q.sub.1 when the latter is biased to its saturation range.
To overcome drawbacks of this kind, tied to the parasitic currents, a conventional solution has been that of optimizing the collection efficiency of the carriers in the collector region, so as to maximize the gain of lateral transistors.
For this reason, lateral PNP transistors have been implemented conventionally with their emitter region occupying, in the active area, a central location surrounded by two collector regions.
While being in many ways advantageous, this prior solution has a serious drawback in that it leads to increased values of certain characteristic parameters of the lateral PNP device, such as the base-collector and emitter-base capacitances, C.sub.bc and C.sub.eb, which restrict performance at high frequencies.
This drawback is also connected with the considerable spread of the base region in the buried layer brought about by the need to control the lateral PNP device performance in the emitter and collector regions.
Especially the width W.sub.b of this base region adversely affects the carrying parameter B*, which is tied to W.sub.b by the following formula: ##EQU1##
where D is the diffusion value, and .tau. is the re-combination time of the carriers.
It should be noted that the B* factor is inversely proportional to the passage time through the base region, so that as it increases, the frequency performance of the lateral PNP device deteriorates.
Unfortunately, this type of lateral PNP device revealed serious limitations when attempts were made to improve its high frequency performance.
These limitations come from the practical impossibility of bringing the collector regions closer to the central emitter region.
This is both attributable to the photolithographic masks used for transferring the patterns of the active region, and side diffusion effects during the formation of the emitter and collector regions, as well as to breakdown effects which may occur in the region between the base and the collector.
As a result, the width W.sub.b of the base region of the lateral PNP transistors always exhibits values between 2 .mu.m and 4 .mu.m.
Another problem to be taken into account is that the central emitter region includes neighboring regions with a predetermined width E.sub.c, due to the "bird's beak" phenomenon.
These regions cause the minimum width W.sub.AA of the active area required for containing the central emitter region (see example in FIG. 7) to be increased, which further harms the device performance.
Thus, it has become good designing practice to never put an active area minimum width W.sub.AA at anything below the limit given by the following formula (1.2), taking account of the actual width of the emitter contact region CNTWidth and of the added width E.sub.C due to the "bird's beak". EQU W&gt;CNTWidth+2*Ec (1.2)
To this regard, the electron microphotograph of FIG. 3 comes useful which shows an enlarged cross-section through a typical lateral PNP device taken near the central emitter region.
In the present instance, it can be seen that the "bird's beak" inside the active area extends for about 0.6 .mu.m and, therefore, requires a minimum width W.sub.AA greater than about 2.6 .mu.m.